Proactive Event-Driven Computing is a new paradigm, in which a decision is not made due to explicit users' requests nor is it made as a response to past events. Rather, the decision is autonomously triggered by forecasting future states. Proactive event-driven computing requires a departure from current event-driven architectures to ones capable of handling uncertainty and future events, and real-time decision making. We present a proactive event-driven architecture for Scalable Proactive Event-Driven Decision-making (SPEEDD), which combines these capabilities. The proposed architecture is composed of three main components: complex event processing, real-time decision making, and visualization. This architecture is instantiated by a real use case from the traffic management domain. In the future, the results of actual implementations of the use case will help us revise and refine the proposed architecture.